Allow artificial network latency between DFaaS nodes

[ema-pe]

Since the basic deployment and tests are performed on VMs hosted on the same machine, the current network latency between DFaaS nodes is negligible. To simulate scenarios in which agents forward requests to other nodes, we may introduce artificial network latency for both the DFaaS agents and HAProxy. This can be implemented directly at the VM level using tc (Linux Traffic Control) (man page, Red Had docs), without relying on ufw (disabled on the VMs) or host-level iptables.

This can also enable further research on handling scenarios where DFaaS nodes have different network latencies (see issue #6 and #41).